An evolving NGF–Hoxd1 signaling pathway mediates development of divergent neural circuits in vertebrates

Summary

Species are endowed with unique sensory capabilities encoded by divergent neural circuits. One potential explanation for how divergent circuits have evolved is that conserved extrinsic signals are differentially interpreted by developing neurons of different species to yield unique patterns of axonal connections. Although NGF controls survival, maturation and axonal projections of nociceptors of different vertebrates, whether the NGF signal is differentially transduced in different species to yield unique features of nociceptor circuits is unclear. We identified a species-specific signaling module induced by NGF and mediated by a rapidly evolving Hox transcription factor, Hoxd1. Mice lacking Hoxd1 display altered nociceptor circuitry which resembles that normally found in chicks. Conversely, ectopic expression of Hoxd1 in developing chick nociceptors promotes a pattern of axonal projections reminiscent of the mouse. We propose that conserved growth factors control divergent neuronal transcriptional events which mediate interspecies differences in neural circuits and the behaviors they control.

The assembly of neural circuits is instructed by phylogenetically conserved families of extracellular cues that control neuronal differentiation, axon and dendrite targeting, survival, and synaptogenesis^1–3. Yet, it is unclear how novel neural circuits arise during evolution to encode unique behaviors among different animal species. In a classic example of extrinsic control of neural development, target-derived nerve growth factor (NGF) promotes survival, maturation, and target innervation of developing nociceptors, which are polymodal sensory neurons that innervate the skin and detect pain, temperature, touch, and itch⁴. A similar, central role of NGF in nociceptor development is highly conserved in different classes of vertebrates, including mammals and birds^5–7. However, mammalian and avian nociceptor circuits display notable differences that likely represent beneficial evolutionary adaptations^8–9. In mammals, for example, nociceptors form elaborate axonal endings associated with hair follicles and within the epidermis^10–11. Such epidermal and hair-follicle-associated nociceptor endings are absent in birds and reptiles^12–14 and may represent adaptive features of mammals since their skin lacks rigid mechanical barriers, such as scales and feathers present in birds and reptiles, and is thus more directly exposed to environmental stressors^15–17. The molecular driving forces behind the evolution of neural circuits underlying mammalian nociception, as for those mediating novel sensory, motor and cognitive functions, remain poorly understood. Here, we test the idea that the conserved NGF signal triggers species-specific patterns of nociceptor gene transcription which underlie divergent organization of vertebrate nociceptor circuits.

Results

A screen for genes controlled by NGF signaling in mammalian nociceptors

We first searched for evolutionarily dynamic transcriptional targets of NGF signaling pathways that may function in mammals, but not in birds, to control unique aspects of nociceptor development. NGF-dependent genes in developing mouse nociceptors were identified through three separate genome-wide screens: (1) One screen compared gene expression profiles of embryonic day (e)14.5 dorsal root ganglia (DRG) of wild-type (WT) and Ngf^−/− mice to identify genes whose transcripts are enriched in nociceptors¹⁸; (2) A second screen compared e14.5 DRGs of Bax^−/− and Ngf^−/−; Bax^−/− mice to identify NGF-regulated genes expressed in nociceptors in vivo (Bax^−/− prevents nociceptor cell death accompanying genetic deletion of Ngf ¹⁹.); (3) The third screen compared mouse DRG explants grown in the presence or absence of NGF for identification of NGF-dependent genes expressed in nociceptors in vitro. These three screens identified a large cohort of NGF-dependent genes in mouse nociceptors, including the well-characterized nociceptor genes Ret, CGRP, and TrkA (table S1–S3). Moreover, a cross-comparison of the three screens revealed a small number of genes that are tightly regulated by NGF both in vivo and in vitro and are potentially important for development of mouse nociceptor circuitry (Fig. 1a, b). We next used parallel cultures of mouse and chick DRGs and quantitative PCR (qPCR) analyses to determine the extent to which the avian orthologs of these identified mammalian genes are also expressed in an NGF-responsive manner in chicken nociceptors (Supplemental Fig. 1). Several observations led us to focus on one gene, which encodes the homeobox transcription factor Hoxd1: (1) All three genome-wide screens identified Hoxd1 as one of the most NGF-responsive genes in mice (Fig. 1b and Table S1–S3); (2) In our cross-species comparison, Hoxd1 was robustly induced by NGF in mouse but not chick nociceptors, in stark contrast to the majority of other NGF-controlled genes, which behaved similarly in mouse and chick nociceptors (Fig. 1c, e and Supplemental Fig. 1); (3) While most Hox genes encode highly conserved transcription factors critical for animal development²⁰, sequence comparisons suggest a high degree of divergence of Hoxd1 between different vertebrate species (Supplemental Fig. 1 and ref. 21). Furthermore, the well-characterized expression of Hoxd1 in the anterior neuroectoderm of lower vertebrates is absent in modern mammals^22–23, raising the possibility of a newly evolved and yet unknown function of Hoxd1 in the mammalian lineage.

Figure 1. NGF signaling induces Hoxd1 expression in mammalian, but not avian, embryonic nociceptors.

(a) Diagram of the numbers of unique genes identified by the three microarray screens (at a loose cutoff of probability > 0.5 and fold change > 1.5). (b) The degree of mRNA level changes of the 14 genes identified by all three independent microarray screens, in log scale. (c) Application of NGF to cultured mouse DRG explants robustly induces transcription of Hoxd1 within 24 hrs. (d) RT-PCR reactions comparing levels of Hoxd1 mRNA in DRGs dissected from e14.5 Bax^−/− and Ngf^−/−; Bax^−/− embryos. (e) In parallel cultures of mouse and chick DRG explants, NGF induces robust transcription of canonical NGF-dependent genes (CGRP and Ret) in both species, but induces Hoxd1 in mouse but not in chick DRGs as assessed by quantitative PCR. * p < 0.05 by Student’s t-test (n = 3 independent batches of cultures for each condition). (f) In situ hybridization for Hoxd1 mRNA in sections of developing DRGs of e14.5 Bax^−/− and Ngf^−/−; Bax^−/− embryos. Scale bar = 50 μm. (g) Comparison of Hoxd1 expression in DRGs and spinal cord of e15.5 mouse and stage 32 (e7.5) chick embryos. TrkA is a general marker for embryonic nociceptors. In chick DRGs, TrkA⁺ nociceptors and TrkA non-nociceptors segregate into distinct topographic domains. sc, spinal cord. drg, dorsal root ganglion. Scale bar = 50 μm and 100 μm in low- and high-magnification panels, respectively. (h) Double fluorescent labeling of Hoxd1 mRNA and TrkA protein in e15.5 mouse and stage 32 chick DRGs. Scale bar = 20 μm.

Developmental expression of Hoxd1 in different vertebrate species

To begin to characterize potential roles of Hoxd1 in development and evolution of vertebrate nociceptor circuits, we first examined its pattern of expression in mouse and chick embryos. Hoxd1 expression was readily detected in ~80% of developing nociceptors of WT mouse embryos (Fig. 1g, h) where its time of onset and peak expression are coincident with that of NGF signaling (Supplemental Fig. 2). Moreover, expression of Hoxd1 was eliminated in genetically modified mice lacking either nociceptive neurons (Ngf^−/−) or NGF/TrkA signaling (Ngf^−/−; Bax^−/−), indicating the presence of an NGF–Hoxd1 signaling module in developing mouse nociceptors (Fig. 1d, f). Furthermore, in WT mouse embryos, nociceptors in DRGs located throughout the rostrocaudal axis express Hoxd1 at comparable levels, a pattern that is atypical for a Hox gene and yet consistent with the aforementioned NGF-dependent expression of Hoxd1 (Supplemental Fig. 2). In contrast and reflecting a lack of transcriptional activation of avian Hoxd1 in response to NGF (Fig. 1e), Hoxd1 mRNA was undetectable in developing chick nociceptors both before and after the onset of NGF signaling in chick embryos (Fig. 1g, h). To determine whether the NGF–Hoxd1 signaling module is present in other non-mammalian vertebrates, we examined Hoxd1 expression in developing DRGs of the reptile green anole lizard (Anolis carolinesis) and the amphibian clawed frog (Xenopus tropicalis). Few, if any, developing nociceptors of these animals express Hoxd1, resembling the lack of expression in chicks (Supplemental Fig. 3). Moreover, an alignment of vertebrate genomes revealed clusters of densely packed sequence motifs surrounding the Hoxd1 transcription unit that are highly conserved in different mammals but absent in non-mammalian genomes (Supplemental Fig. 3). Thus, while peripheral target-derived NGF supports nociceptor survival and axon extension in rodents, birds and perhaps other vertebrate species^4–7, it promotes robust expression of Hoxd1 in mouse but not chicken nociceptors.

Nociceptive circuitry in the skin of Hoxd1^−/− mice resembles that of non-mammalian vertebrates

To determine whether Hoxd1 mediates development of divergent nociceptor circuits in vivo, we next generated mice harboring a targeted deletion of Hoxd1 (Supplemental Fig. 4). Hoxd1^−/−mice are viable, fertile and, unlike mice harboring mutations in other anterior group Hox genes, exhibit normal hindbrain patterning (data not shown). Intriguingly, however, the pattern of nociceptor innervation of the skin of Hoxd1^−/− mice is abnormal and reminiscent of that which is normally observed in non-mammalian vertebrates, including birds. Our initial focus was on sensory innervation of hair follicles because hair is a distinguishing feature of mammals that represents a cold adaptation feature necessary for the prevention of heat loss¹⁶. Moreover, in mammalian skin, hair follicles serve as organizing centers for axons of DRG somatosensory neurons, including nociceptors¹⁰. In WT mice, nociceptor endings, which are strongly labeled by anti-Peripherin staining, encircle most if not all down hair follicles (Fig. 2a, b). Birds, on the other hand, have evolved feathers, and feather follicles do not possess such circular sensory endings (Fig. 2a and ref. 12). Interestingly, Hoxd1 mutant mice exhibit a marked deficit of the Peripherin⁺ axonal endings associated with hair follicles (Fig. 2b, e). A key role of Hoxd1 in nociceptor hair afferent development is further supported by the observation that it is required for specification of Mrgprb4⁺ neurons, which constitute a unique population of mammalian nociceptors that exclusively innervate hair follicles²⁴. In mice, Mrgprb4⁺ neurons represent ~15% of non-peptidergic nociceptors, and their peripheral axons branch exclusively in hairy skin and encircle the neck of hair follicles²⁴. In Hoxd1^−/− mice, there are ~3 times more DRG sensory neurons expressing Mrgprb4 (Fig. 2d, g). Moreover, all Mrgprb4⁺ neurons in Hoxd1^−/− mice abnormally express markers of peptidergic nociceptors, CGRP and TrkA, indicating impaired specification of these hair follicle-associated nociceptors (Fig. 2d, f). This nociceptor specification deficit appears to be restricted to hairy skin nociceptors as expression of molecular markers for the major subclasses of nociceptors is largely normal in Hoxd1^−/− mice (Supplemental Fig. 5). Lastly, in mammals, the epidermis of non-hairy (glabrous) skin is innervated by a dense array of nociceptor endings that convey noxious chemical, mechanical, and temperature stimuli^10–11. Such epidermal free nerve endings are absent in birds and reptiles despite the presence of dermal fibers^13–14,25, as the epidermis of these species is directly protected by mechanical barriers (scales and feathers) composed of a unique tough keratin protein that is not produced by mammals^15–17. Strikingly, in Hoxd1^−/− mice, while neuronal number and dermal fibers are present in normal abundance, there is a reduction of epidermal endings of peptidergic nociceptors (Fig. 2c, h), reminiscent of that which is normally found in the skin of non-mammalian vertebrates. Thus, NGF-dependent expression of Hoxd1 in mammalian nociceptors controls mammal-specific features of nociceptor development, including specification and target innervation of subsets of hair-follicle-associated nociceptors and penetration of the epidermis by peptidergic nociceptors.

Figure 2. Hoxd1 instructs development of mammal-specific features of nociceptor axonal connections in the skin.

(a) Diagram showing the distinct patterns of nociceptor endings in the skin of mammals and birds. (b) Anti-peripherin staining of mouse hairy skin from proximal limb. Upper panels: transverse sections; lower panels: cross sections. (Hair is autofluorescent.) Scale bar = 40 μm and 50 μm in upper and lower panels, respectively. (c) A dramatic reduction in the number of nociceptor free nerve endings that penetrate the epidermis of hindlimb footpad glabrous skin of Hoxd1^−/− mice was detected by anti-CGRP staining, which labels peptidergic nociceptors. Arrows indicate epidermal nociceptor endings. Scale bar = 40 μm. (d) Double fluorescent labeling of Mrgprb4 mRNA and CGRP protein in L3 DRGs of WT and Hoxd1^−/− mice. (e) Quantification of the number (±SEM) of transverse lanceolate endings per mm² skin area in serial sections of hindlimb back thigh hairy skin. * p < 0.001 by Student’s t-test. (f) Percentage (±SEM) of Mrgprb4⁺ neurons in WT and Hoxd1^−/− mice that co-express CGRP or TrkA. * p < 0.001 by Student’s t-test. (g) Quantification of the percentage (±SEM) of lumbar DRG neurons that express Mrgprb4. * p < 0.001 by Student’s t-test. (h) Quantification of the average (±SEM) number of CGRP⁺ free nerve endings crossing the dermal-epidermal boundary per unit length (300 μm) of hindlimb footpad glabrous skin. * p < 0.005 by Student’s t-test.

Hoxd1 instructs nociceptor central axonal projections within the mammalian spinal cord

Mirroring species-specific differences in innervation of the skin, the patterns of nociceptive axonal projections within the spinal cords of mammals and birds are also distinct^8–9. In mice, the majority of central axonal branches of nociceptors penetrate the dorsal spinal cord in the dorsal region, project ventrally, and terminate in lamina I and II of the dorsal horn (Supplemental Fig. 6). In contrast, avian (chick) nociceptors penetrate the spinal cord from a more lateral position and, in addition to those nociceptors that terminate in lamina I and II of the dorsal horn, many nociceptor axons extend horizontally into deep layers of the spinal cord⁹ to innervate integrative regions that receive a convergence of sensory inputs of multiple modalities^26–27 (Supplemental Fig. 7). Remarkably, in Hoxd1^−/− mice, we detected a large number of aberrant nociceptor axonal projections within the spinal cord that bear intriguing similarity to the horizontally-extending deep projections normally observed in birds (Fig. 3a, b). The aberrant nociceptor axons enter the spinal cord of Hoxd1^−/− mice from a lateral position, project horizontally across the spinal cord, and enter deep regions near the central canal (Fig. 3a). These misrouted fibers are detected at all axial levels (Fig. 3c, d) and terminate in and around lamina X, an area purported to integrate somatic and visceral sensory information of multiple modalities^26–27. Although such deep fibers were rarely observed in WT mice, the few that were seen were largely restricted to lumbrosacral levels (L4-S1; Supplemental Fig. 6) and are likely to be visceral organ afferents^28–29. In further support of a role of NGF-dependent Hoxd1 expression in directing nociceptor connections within the mouse spinal cord, mutant mice deficient for NGF signaling (Ngf^−/−; Bax^−/−) similarly exhibit excessive nociceptor axons within deep spinal cord layers (Fig. 4). Moreover, consistent with this, lizards, which do not express Hoxd1 in nociceptors (Supplemental Fig. 3), exhibit deep-projecting nociceptor fibers similar to those in chicks and Hoxd1^−/− mice³⁰. These observations suggest that Hoxd1 mediates NGF-dependent suppression of nociceptor projections into deep layers of the spinal cord.

Figure 3. Hoxd1 controls a species-specific pattern of nociceptor axonal connectivity within the spinal cord.

(a, b) Anti-CGRP staining of the dorsal spinal cord of P0 mice and stage 39 (e13) chick. (a) Deep nociceptor central projections are abnormally increased in the spinal cord of Hoxd1^−/− compared to WT mice revealed by anti-CGRP, a peptidergic nociceptor marker. More numerous deep projections were found at all axial levels, including cervical (Cer) and thoracic (Th) segments. V, lamina V; X, lamina X. Arrows point to aberrant deep projections extending into a region near the central canal (lamina X). Some CGRP⁺ axons normally reach lamina V similarly in WT and Hoxd1^−/− mice mostly from ventral projections emanating from the dorsal funiculus. (b) Anti-CGRP staining of nociceptor central projections in stage 39 (e13) chick spinal cord at lower cervical and thoracic levels. Arrows indicate prominent horizontally-projecting deep nociceptor axon bundles in the spinal cord of the chick that resemble the abnormally increased fibers in Hoxd1^−/− mice. Arrowhead indicates axon bundles that first project ventrally and then turn medially. Scale bar = 50 μm in (a) and (b). (c) Quantification of (a) comparing the fraction of 20μm sections of the spinal cord at different axial levels that contains bundles of deep nociceptor axons in WT (n = 5 animals) and Hoxd1^−/− (n = 5 animals) mice. Numbers denote the numbers of sections examined. (d) Quantification of (a) comparing the average (±SEM) numbers of individual deep CGRP⁺ fibers detected per section at different axial levels * p < 0.05 by Student’s t-test. (n = 5 animals in each group.)

Figure 4. NGF signaling controls a subset of central axonal projections of mammalian nociceptors in the spinal cord.

(a) Aberrant nociceptor axons, labeled with anti-Peripherin, project horizontally into deep regions of the spinal cord in Ngf^−/−; Bax^−/− mice. Lower panels represent high-magnification of the images shown in upper panels. Dashed boxes indicate excessive deep nociceptor axons. cc, the central canal. (Inset) Peripherin is expressed at high levels in nociceptors^19,38. Note that in Ngf^−/−; Bax^−/−mice, in addition to excessive horizontal projections, there is also an abnormally large number of aberrant deep nociceptor axons that project ventrally from the medial edge of the dorsal horn (arrowhead), some of which extend beyond the central canal. Scale bar = 60 μm and 30 μm in low- and high-magnification panels, respectively. (b) Diagram representing nociceptor central projection defects of Ngf^−/−; Bax^−/− mice in comparison to Bax^−/− controls. (c, d) Quantification of (a). (c) Fluorescent intensities are measured from areas represented by dashed rectangular boxes in (a). Average (±SEM) fluorescent intensities of Peripherin⁺ axons in deep ventral and horizontal projections are shown. * p ≤ 0.001 by Student’s t-test (n ≥ 5 animals in each group). (d) Fraction (%) of level-matched lumbar spinal cord sections that show anti-Peripherin fluorescent intensity in horizontal projections.

Ectopic Hoxd1 expression in chick nociceptors induces mammal-like traits

If NGF-dependent expression of Hoxd1 is indeed a determinant of the mammal-specific pattern of nociceptor circuits, then ectopic expression of Hoxd1 in non-mammalian nociceptors may alter their axonal projections to adopt mammal-like features. Therefore, we asked whether ectopic Hoxd1 expression in nociceptors of the chick is sufficient to alter their pattern of axonal projections within the spinal cord. Using a neural crest-specific driver³¹, Hoxd1 was ectopically expressed in chick DRGs on one side of the animal, leaving the other side as a control (Fig. 5a, b), an approach that enables a direct comparison of sensory afferent projections between the electroporated and unelectroporated sides of the spinal cord³². Remarkably, ectopic expression of chick Hoxd1 affected the pattern of chick nociceptor central projections (Fig. 5c–e). In control embryos, at stage 35, TrkA⁺ nociceptor central axons have penetrated the superficial dorsal horn at a lateral position and have projected into deep layers of the dorsal horn. Many of these axons that initially project horizontally then turn in a ventral direction and form prominent bundles of horizontally-projecting axons (Fig. 5c and Supplemental Fig. 8). Following ectopic expression of Hoxd1 in chick DRGs, however, the prominent deep nociceptor projections, which under normal conditions extend horizontally toward the central canal, are suppressed (Fig. 5c). Importantly, electroporation of a Hoxd1 construct lacking the DNA-binding homeobox motif (Hoxd1Δh) had no effect on nociceptor projections (Fig. 5d). Moreover, the central projections of proprioceptive neurons appear normal following ectopic Hoxd1 expression (Supplemental Fig. 8). Thus, ectopic expression of Hoxd1 alters the central projection programs of chick nociceptors, forcing them to adopt murine nociceptor features (Fig. 5e).

Figure 5. Ectopic expression of Hoxd1 in chick nociceptors impairs their axonal ingrowth into the lateral spinal cord.

(a) DNA constructs used for in ovo electroporation. (b) Sox10 enhancer-driven in ovo electroporation directs ectopic gene expression in developing chicken DRGs but not in the spinal cord. Left panel: A whole-mount lateral view of GFP fluorescence in stage 26 chick embryos 3 days after electroporation. Right panel: Transverse sections of embryos processed as in the left panel, stained with anti-GFP. sp, spinal cord. drg, dorsal root ganglion. (c) The pattern of central projections of chick nociceptors is altered when Hoxd1 is expressed in the chick DRG. Upper panel: A spinal cord section labeled with anti-TrkA at stage 35 (e9); Lower panels: high-magnification views of the boxed areas of the same spinal cord section. The left side is control, and DRGs of the right side of the embryo are electroporated with Hoxd1. Scale bar = 50 μm and 25 μm in low- and high-magnification panels, respectively. (d) Quantification of (c) shows the ratio between spinal cord areas occupied by TrkA⁺ fibers in the electroporated side versus the control side. * p = 0.0001 by Student’s t-test (n > 4 embryos for each group; sections are 25 μm thick and are sampled at least 200 μm apart). cHoxd1, chicken Hoxd1 gene. cΔ, a chicken Hoxd1 gene construct lacking the homeobox motif. mHoxd1, mouse Hoxd1 gene. mΔ, a mouse Hoxd1 construct lacking homeobox. (e) Diagram of the different patterns of spinal cord innervation by mammalian and avian nociceptors.

Hoxd1^−/− mice have deficits in pain sensitivity

To assess the behavioral consequences that may result from aberrant nociceptor circuits in Hoxd1^−/− mice, we examined the Hoxd1^−/− mutants for potential defects in somatosensation. Adaptation to cold climates is a central theme of mammalian evolution³³, and many evolutionarily novel features of mammals are related to cold adaptation³⁴. Indeed, the ancestors of modern mammals survived and flourished during extended geological periods of cold temperatures when many terrestrial heterothermic species became extinct^35–36. Remarkably, using measurements of avoidance of extreme hot and cold temperatures, we found that Hoxd1^−/− mice exhibit markedly longer latencies of avoidance to an extremely cold (0°C) surface (Fig. 6a), representing a defective self-protection behavior essential for the survival of animals that explore cold environments³⁷. Hoxd1^−/− mice are otherwise comparable to littermate controls in their responses to hot temperature and light mechanical stimulation (Fig. 6b–e), although they do show abnormally prolonged carrageenan-induced thermal allodynia (Fig. 6f), suggesting a role of Hoxd1 in inflammatory pain. The causes of the behavioral deficits of Hoxd1^−/− mice are uncertain; Such deficits may arise from defects in axonal wiring in the skin, the spinal cord, or in defective specification of nociceptive neurons in the mutant mice. These results indicate that Hoxd1 is essential for the development of behavioral responses to cold temperatures in mice.

Figure 6. Behavioral responses of Hoxd1^−/− mice to somatosensory stimuli.

(a) Hoxd1^−/− mice show a defect in their avoidance response to extreme cold. The paw licking or flinching response latency following exposure of mice to a 0°C cold plate is significantly increased in Hoxd1^−/− mice compared to their WT littermate controls (n = 14 for WT, 13 for KO). *** p < 0.001 by Student’s t-test. (b–d) Hoxd1^−/− mice respond normally to acute noxious thermal stimuli. Response latencies in (b) the Hargreaves (n = 12 for WT, 11 for KO), (c) hot-plate (50°C; n = 9 per genotype) and (d) tail-immersion (50°C; n = 9 per genotype) tests do not differ between WT and Hoxd1^−/− mice. (e) The paw withdrawal threshold of Hoxd1^−/− mice to punctate mechanical stimuli (von Frey test) is comparable to that of WT mice (n = 12 per genotype). (f) Hoxd1^−/− mice show prolonged thermal hyperalgesia 24 hours after intraplantar injection of 1% carrageenan as compared with WT mice (10 μl; n = 9 per genotype). Pre, preinjection. Data are presented as Mean (±SEM). * p < 0.01 Two-way ANOVA.

Discussion

Here, we report that Hoxd1 is a rapidly evolving NGF/TrkA signaling pathway component whose differential expression in developing nociceptors across vertebrate species contributes to species-specific features of nociceptor circuits. We identified Hoxd1 from a genome-wide screen for genes expressed in response to NGF signaling in mice but not chickens. Moreover, genetic manipulations in mammals and birds showed that Hoxd1 instructs development of mammal-specific features of nociceptive neural circuitry. Finally, behavioral sensitivity to extreme cold is markedly compromised in Hoxd1 mutant mice, supporting functional significance of acquisition of Hoxd1 by mammalian nociceptors.

The Hox genes have undergone extraordinary evolutionary selection to delineate variations of body and limb plans in divergent lineages across the animal kingdom²⁰. The present study presents a particularly fascinating feature of Hox gene divergence such that in the mammalian lineage the conserved axial patterning functions of Hoxd1 are lost and replaced by a role in mediating development of nociceptor circuits. Redundancy between Hoxa1, Hoxb1 and Hoxd1 may have allowed for relaxation of Hoxd1 from conserved roles and to evolve novel functions in mammals. Loss of Hoxd1 in mice results in compromised responses to extremely cold temperatures, an environmental stress that has challenged terrestrial animals throughout recent geological history^35–36. Repeated, dramatic, and long-term global cooling in the Cenozoic Era may thus represent selective pressure for mammalian evolution of Hoxd1. More generally, our analysis indicates that altered expression of a single, evolutionarily dynamic transcription factor during the course of vertebrate evolution can instruct unique features of sensory circuits and the behaviors they subserve.

We show that mammalian features in nociceptor circuits are subject to coordinated developmental control by mammal-specific expression of Hoxd1. Remarkably, Hoxd1 stands out as one gene that is robustly induced by NGF/TrkA signaling in rodents but not in birds amid a largely conserved genomic transcriptional response to NGF. We propose a model in which divergent expression of one or a small number of genes in embryos of different species, induced by conserved growth factor signaling pathways, results in macroscopic interspecies differences in adult neural circuits. Thus, identification and characterization of differentially expressed effectors of conserved growth factor signaling pathways will likely reveal key determinants of anatomical features, behaviors, and diseases unique to divergent vertebrate organ systems.

Methods

Generation of Mice and Mouse Genetics

Hoxd1^−/− mice were generated by a targeting strategy similar to those described for several other Hox genes^6–7. A polII-Neo-HPRT-pA cassette was inserted into the second exon of the Hoxd1 locus disrupting most of the homeobox domain including the DNA recognition helix and sequences downstream (Supplemental Fig. 4a). Positive 129/Sv ES cell clones confirmed by Southern hybridization (Supplemental Fig. 4b) were used to generate Hoxd1^+/− founders that were confirmed by Southern blot, in situ hybridization, and PCR of genomic DNA extracted from the tails (Supplemental Fig. 4b–d). Hoxd1^−/− mice were genotyped using a multiplex PCR reaction with the following primers: 5′-TCAGAAAATGCCCCAGGAGC (forward), 5′-TAGCCGTTATTAGTGGAGAGG (mutant reverse) and 5′-TGGAACCAGATTTTGACCTGG (WT reverse). The Ngf^+/− and Bax^+/− alleles have been described previously^8–9. For all of our analyses, the morning after coitus was designated as embryonic day (e)0.5 and the day of birth as postnatal day (P)0.

All experimental procedures involving animals have been approved by the IACUC of Johns Hopkins University.

Microarray Analyses

DRGs were dissected out from e14.5 (1) Ngf^−/−; Bax^−/−, (2) Bax^−/− and (3) Ngf^−/− mouse embryos and (4) their control WT littermates from multiple litters and were stored in RNAlater (Ambion) at −20°C until use. DRG explants that were cultured either with NGF or with NT3 were also collected (see below). Two independent replicates, each consisting of DRGs pooled from multiple embryos of like genotypes from different litters, were separately prepared. At least 5 μg of total RNA was isolated from each group of pooled DRGs using an RNeasy Micro Kit (Qiagen). RNA was then hybridized to Affymetrix GeneChip Mouse Genome 430 2.0 arrays by the JHMI Microarray Core Facility. Statistical analyses were performed essentially as previously described^18,39.

A loose cutoff at a posterior probability > 0.5 (ref. ³⁹) and an estimated mRNA fold change > 1.5 was used to generate the Venn diagram (Fig. 1a). The 14 unique genes identified in all three independent microarray screens are: Aldoc, Crtac1, Dgkh, Hoxd1, Idi1, Klf7, Nrxn1, Pou4f2, Ppp1r, Prkca, Ret, Rgs4, Rgs7 and Tle4. These 14 genes and other select candidate genes were tested by qPCR for NGF-responsiveness in mouse and chick DRG cultures unless the sequences for chicken orthologs were not available.

DRG Neuron Explant Culture

For microarray analysis, DRGs were dissected out from 4 litters of e13.5 mouse embryos, placed on plates coated with poly-D-lysine and laminin, and maintained for 48 hrs with 50 ng/ml of either NGF or NT3. DRGs were then collected, and total RNA was extracted for two independent sets of cultures and processed for microarray analysis.

For expression comparisons, DRG explant culture conditions and protocols were previously described⁴⁰. Briefly, DRGs were dissected from e13.5 mouse embryos and e7.25 chick embryos, placed on coverslips coated with collagen, and cultured with media containing 25 ng/ml NGF for 36 hours. DRG explants were then washed, deprived of NGF by culturing with BAF and anti-NGF antibody for 24 hrs, re-washed, and further cultured with BAF and either with or without NGF for an additional 24 hrs.

Chick, Frog and Reptile Embryos and cDNAs

Chicken (Gallus gallus), clawed frog (Xenopus tropicalis), and green anole lizard (Anolis carolinesis) embryos were collected and staged^1,13–14. Total RNA was extracted from stage 21~22 chicken and stage 52 frog embryos and first-strand cDNA synthesis was performed from the RNA using oligo(dT) primers and SuperScript III system (Invitrogen). A cDNA library of stage 7 Anolis carolinesis embryos was a generous gift from Dr. Thomas Sanger (Harvard University). Lizard and frog TrkA genes are identified based on homology searches using the highly conserved kinase domain, and lizard Hoxd1 was identified by homology and the genomic structure of the HoxD cluster and is in agreement with previous reports⁴¹. PCR products were blunt-end cloned into a pSC-B vector (Stratagene) and used as templates for synthesis of in situ hybridization probes. The following primers were used to amplify fragments of the TrkA and Hoxd1 genes from the cDNA.

• Frog TrkA: 5′-GATAATCCCTTCAGCTTTAACC and 5′-ATCTTGACCACCAGATTATTGC

• Frog Hoxd1: 5′-GAAAGATTGAGGGATCATCAGG and 5′-TCGAGTTCAGTCAGTTGTTTGG

• Lizard TrkA: 5′-CACCGAAGAGAAGATGCTGGT and 5′-TAAATACACCGGGGGAGCTTT

• Lizard Hoxd1: 5′-AAGCAGAAGAAAAGGGAGAGAGA and 5′-TGTAAGTGAAGTTGGGGAGAGAA

Immunohistochemistry and In Situ Hybridization

Protocols for immunohistochemistry and in situ hybridization were essentially as described³⁸. For immunohistochemistry, mouse tissues were either preserved directly in OCT (Tissue Tek), fixed or transcardially perfused with either 4% ice-cold paraformylaldehyde (PFA) or Zamboni’s fixative then cryoprotected in 30% sucrose/phosphate-buffered saline (PBS). Immunofluorescent stainings were performed with cryosections either on glass slides or as floating sections. Primary antibodies used include: rabbit anti-mouse TrkA (Millipore, 1:1000), rabbit anti-chicken TrkA and TrkC (gifts from Dr. Frances Lefcort, Montana State University, 1:1000), rabbit (Millipore, 1:1000) or mouse anti-peripherin (Millipore, 1:500), rabbit anti-CGRP (Millipore, 1:1000), rabbit anti-PGP9.5 (Millipore, 1:1000), rabbit anti-Parvalbumin (Swant, 1:2000), rabbit anti-GFP (1:2000, Invitrogen), chicken anti-LacZ (Promega, 1:500), rabbit anti-Ret (Immuno-Biological Laboratories, 1:1000), mouse anti-chick Neurofilament 160kD (Zymed, 1:500) and rabbit anti-Neurofilament 200 (Millipore, 1:1000). Secondary antibody incubations were performed with Alexa Fluor-conjugated secondary antibodies (Molecular Probes, 1:500).

Probes used for in situ hybridization were: mouse Hoxd1 (nucleotides 946–1386 of NM_010467.2), mouse CRD-Nrg1 (nucleotides 1–598 of NM_178591.2), rat Nrp1 (nucleotides 2515–3799 of NM_145098), chick Hoxd1 (BU269116, (clone ChEST816c12, GeneServices) digested by PstI and transcribed with T3) and frog Hoxd1 (nucleotides 7–701 of NM_001016678.2). Sequences of in situ probes for lizard Hoxd1 and lizard TrkA are described above in the “Chick, Frog and Reptile Embryos and cDNAs” section.

Imaging, Quantification and Statistical Analyses

Confocal fluorescent images were taken on a Zeiss Axioskop 2 microscope with an LSM 510 confocal scanning module. Whole-mount fluorescent images were taken from a Zeiss SteREO Lumar.V12 fluorescent dissecting microscope. To quantify hairy skin innervation, an approximately 4 mm X 4 mm piece of the back thigh hairy skin was dissected from young adult animals (P20 – P30), and serial cryosection was performed at a thickness of 25 μm. Every third or fifth section was collected, and the total number of circular hair follicle endings was counted in each of the collected sections. The number was then normalized to the area of the skin measured in ImageJ. The same skin region was taken from each animal. Quantification of epidermal free nerve endings was performed as described³⁸. Fluorescent intensities were measured using ImageJ (NIH), and a rectangular measuring window with the same size was applied to all images that were compared. The statistical analyses of significance, including p value cutoffs, are specified in Figure Legends.

Quantitative RT-PCR

First-strand cDNA was synthesized from 1 μg of total tissue RNA using oligo(dT) primers and SuperScript III First Strand Synthesis System (Invitrogen). Real-time PCR (qPCR) was performed using an Applied Biosystems 7300 Real-Time PCR System using Quantitect SYBR Green PCR Kit (Qiagen). Expression levels for each gene from each sample were normalized to control PCR products of Gapdh or Actb for non-neuronal tissues, or Pgp9.5 (alias Uchl1) specifically for neurons¹⁸. Primers for qPCR were:

• Mouse Hoxd1: 5′-CCCCAAGAAAAGCAAACTGTCC and 5′-AAACAGTTGGCTATCTCGATGC;

• Chicken Hoxd1: 5′-AGAGGAGGAGCAGAAACCCCAG and 5′-TTTGCACAGGAAGGGAGGAAGAC;

• Mouse Ret: 5′-AGAAAAGCAAGGGCCGGATTC and 5′-GTCGTTCAGGAGGAATTCCAGG;

• Chicken Ret: 5′-AAAGTGATGTGTGGTCGTTTGG and 5′-CACTGCAGTTTTCTGGTCTCTCC;

• Mouse CGRP: 5′-AGGGCTCTAGTGTCACTGCTCAG and 5′-CACATTGGTGGGAACAAAGTTG;

• Chicken CGRP: 5′-CAGTCAGACCTGGCTTGGAGTC and 5′-TGTTCCCCTCAGAGGCTTGCTC;

• Mouse Klf7: 5′ - GAGCTCGGGGGAGGGTTTCTCC and 5′ - GCCGGTGTCGTGGACAAGTTGT

• Chicken Klf7: 5′ – TGGCAGCAGACATGCCTCGAGT and 5′ - GGGGTCCAGCCGTCGGATACTG

• Mouse Ldb2: 5′ - GCACGTCCAATAGCAGTGCCGG and 5′ - TCCTACCACCATCACATCAGGTACCT

• Chicken Ldb2: 5′ - TCCAGCGCACCACATGACCCTT and 5′ - GTTGTCACTGTCCTCGCTCCGC

• Mouse Diras2: 5′ - GGACCGTGAGCCTCCAGATCGA and 5′ - TCCTGCACGAAGGCCTCTCACA

• Chicken Diras2: 5′ - GTGGTTGTGTTTGGGGCTGGAGG and 5′ - TGCCGGTAGGTGTCTTCAATGGTG

• Mouse Ncam2: 5′ - GCAAAGTAGAGCTTAGTGTGGGTGAG and 5′ - GTCGTGACCTGACACCCTCCTT

• Chicken Ncam2: 5′ - AGCAAAGTCGAACTCAGTGTGGGT and 5′ - ACGTGACCGAACACCTTCTTTTTGA

• Mouse Nrxn1: 5′ - AGAAGGAAGCCGTGTTGGTGCG and 5′ - CGATGGCGATGTCATCTGTCCCA

• Chicken Nrxn1: 5′ - TCGTACAACTGAGCCAAATGGTTTGA and 5′ – ACAGCAGGTAGAGGTGGCCATCA

• Mouse Runx1: 5′ - CTCCAGCCTCCAGCTACTGCCC and 5′ - CCTGAGCCTGACAGTGCCCCTT

• Chicken Runx1: 5′ - TGGAAGAGGGAAAAGCTTCACAC and 5′ – CTTGCCTCGTGTTTCTGGGTTC

• Mouse Rgs7: 5′ - ACGGAGCCATCGCCCATACCTT and 5′ - ACAGCGTCCCTTGGCAAGTGTG

• Chicken Rgs7: 5′ - GCTGGTGTACAGAAAGATGGAAGACG and 5′ -TGTCTGAACCAGAAAAGACACTGGGG

• Mouse PlxnC1: 5′ – TGCCTGGCCCATCTAGACTGCA and 5′ - ACAAGATTGGGAGGGCGTCTGT

• Chicken PlxnC1: 5′ – TGCAGTGTCCAAAATGCTTCCAAGG and 5′ -TGCCCAACGTAGAAATCCACAAAGGC

• Mouse Ppp1r: 5′ - AGCATGCTTCTGTGGACTCCCCA and 5′ - AGGTTTAGCACCCACCTCCATGGT

• Chicken Ppp1r: 5′ – GAGTGCCTGTGCTACTCGCACC and 5′ - CCCCAAGTGAGCAGATCACCAGC

• Mouse Prkca: 5′ - CCCACAGCGGGATTTTCCCGTG and 5′ - CCGTCATCCCCACAGTTGAGCG

• Chicken Prkca: 5′ – GGGACAGTTGGTGAGCGAACCG and 5′ - AAATCCCAGGCGCCAAAGGACG

• Mouse Crtac1: 5′ – TGATGGGGGTTCCGGCTACCTG and 5′ -GGCCACACTTCGGCTCACCATC

• Chicken Crtac1: 5′ - CGAGAACGGACGCTGCGTAGAC and 5′ - AGCGGTACCCGCCATAGGTGTT

• Mouse Etv4: 5′ – ATGCAGAAGGTGGCTGGCGAAC and 5′ - TGTCCTCCTCACTGACTGGCCG

• Chicken Etv4: 5′ - GCTGTGCCATGTGGCTGCAAAC and 5′ - CAGGCAAGGAGGTGGCAATGCA

• Mouse Rgs4: 5′ - GCCAATGTACCGGGCTGCAAAA and 5′ - GATGGGGGAGCTCTGGGGACAT

• Chicken Rgs4: 5′ – CTCGTTCCTGCTCCTGGGCTCT and 5′ - TGCCGTGTCAGAACCCCTGTGA

• Mouse Tle4: 5′ - AGGACTGCTTGCCTACCGCTGT and 5′ - TAAAGCAAGCCACCATCCCGGG

• Chicken Tle4: 5′ - AGTACATTGTCACTGGCTCTGGGG and 5′ - AGTGCTAGGAGAAGGAGTTCTGCCT

• Mouse Dgkh: 5′ - AGCCATCAACGTCAAGGCGCTC and 5′ - TGTAAGGCACTGGACACCCGCT

• Chicken Dgkh: 5′ – ACTGCAGCACCACCGGATAGCT and 5′ - CTCCTGGGGGCTGGATCCATGC

• Mouse Pou4f2: 5′ - CGGCACACGTGCTCTCTCACTC and 5′ - TGGAGACTGCACCTAAGGCTGGG

• Chicken Pou4f2: 5′ - ACTCTTGGAGCACCTGACGCCG and 5′ - ATGGGGTTCATGCCGGCCATG

• Chicken Gapdh: 5′-GAAGGCTGGGGCTCATCTGAAG and 5′-CACGATGCATTGCTGACAATTTTC;

• Mouse Gapdh: 5′-ACTGTGGTCATGAGCCCTTC and 5′-CATGTTTGTGATGGGTGTGAAC;

• Mouse Pgp9.5: 5′-ACGGCCCAGCATGAAAACT and 5′-GACGGATCCATCCTCAAATTCC;

• Mouse Actb: 5′-CCACTGCCGCATCCTCTTCC and 5′-CTCGTTGCCAATAGTGATGACCTG.

Chick In Ovo Electroporation

Chick in ovo electroporation was performed as described^42–43. Briefly, DNA was injected into the neural tube of Hamburger & Hamilton (HH) stage 12~13 chick embryos using a glass pipette. Electrical field was then applied with a pair of needle electrodes (Genetrodes; Harvard Apparatus) and a square pulse electroporator (BTX) at 30 V (pulse length: 50 ms; number of pulses: 5; intervals: 1 s). Successfully electroporated eggs, visually confirmed by endogenous GFP signal under a Zeiss SteREO Lumar.V12 fluorescent dissecting microscope 2 or 3 days after electroporation, were incubated until stage 34~35 for analysis. Chicken Hoxd1 coding sequence was amplified from a chicken cDNA library with: 5′-ATGGACTTGGGTTTTTTCCC (Forward) and 5′-TCAATACAAAGGGAGTAAAGCAGC (Reverse). Truncated chick Hoxd1 lacking homeobox was amplified with the same forward primer and 5′-CTGCCGGGTGCTGAAACTG (Reverse).

To achieve selective expression in the DRGs, we used a strategy similar to what has been demonstrated for motor neuron-specific electroporation¹⁸ but instead searched for enhancers that direct DRG-specificity. Briefly, two constructs were co-electroporated: the first was electroporated at a lower concentration (0.2 μg/μl) and uses an enhancer connected to a weak basal promoter (CMV) to drive Cre recombinase expression. The second construct, electroporated at a high concentration (1~2 μg/μl), uses a strong promoter (CAGGS) to drive a loxP-STOP-loxP-GeneX-IRES-eGFPf cassette. Over 10 enhancer fragments that direct DRG-specific expression in transgenic mice were tested, and one fragment of the Sox10 enhancer (“D6”, described in ref. ³¹) was used. Primers used to clone this fragment from mouse genomic DNA were: 5′-ATGACCGTAAGCCTTAGGAGATGG (Forward) and 5′-GCACTGGCATAGTTGGAAGAATGG (Reverse).

Pain Behavioral Tests

Pain behavioral tests were done essentially as described⁴⁴. Briefly, for the Hargreaves’ test (radiant heat), mice were placed in plexiglas restrainers and a light beam was directed to the footpads of hindpaws using an IITC Plantar Analgesia Meter. For tail immersion, mice were gently restrained in a specially machined 50 ml conical tube that permits ventilation, and the distal halves of their tails were immersed in a 50 °C water bath. For the hot plate test, mice were placed in an IITC Hot Plate Analgesia Meter, and the onset of brisk paw lifts and/or flicking/licking of the paw was assessed. For cold plate, a clear Plexiglas cylinder was placed on a metallic plate equilibrated to 0 °C, and mice were placed inside the cylinder. The latency to the onset of brisk and persistent paw withdrawal and/or licking of the paw was measured. For the von Frey assay, mice were placed within Plexiglas restrainers on a fine metallic grid, and their hindpaws were stimulated using a set of von Frey filaments (Bioseb) starting with the lowest force. For carrageenan-induced inflammation, 10 μl of 1 % carrageenan (Sigma 22049) dissolved in sterile saline was injected subcutaneously into right hindpaws using a Hamilton syringe, and thermal hyperalgesia was measured using a Hargreaves’ apparatus before, 6 hrs, and 24 hrs after the injection. All mice used for behavioral studies were 2~3 month old and were backcrossed to the 129/Sv mouse strain for at least 4 generations.